Apple May Use Nvidia Blackwell in Google Cloud for Siri Overhaul

Headline: Apple may tap Nvidia’s Blackwell chips in Google data centers for stalled Siri overhaul — a win for confidential compute, a headache for decentralization Apple is reportedly preparing to route the long-delayed, cloud-powered version of Siri through Google data centers running Nvidia’s Blackwell B200 processors after efforts to run the updated assistant on Apple’s own Private Cloud Compute (PCC) fell short on speed. What’s changing - According to The Information, Apple tested a modified version of Google’s Gemini model on PCC servers (Apple’s privacy-focused system built around Mac-series chips) but found performance too slow for the real-time experience it wants for Siri. - As a result, Apple is said to plan using Nvidia B200 processors hosted inside Google’s infrastructure to handle the heavy-lifting for cloud-based Siri requests. That would mark a notable shift away from Apple’s long-standing preference for end-to-end control of its hardware, software and cloud stack. Timing and rollout - The move could be revealed at Apple’s Worldwide Developers Conference keynote on June 8, which is expected to include previews of iOS 27 and major Apple Intelligence updates. Yahoo Tech reports the revamped Siri could reach users with the iOS 27 release in September. Privacy and branding questions - Apple has presented PCC as a privacy-first system designed to process user requests on Apple hardware without retaining or exposing data. Reports say Apple may still keep the PCC name even if some processing happens on Google-hosted Nvidia gear. - To address privacy concerns, Apple reportedly plans to use Nvidia’s confidential computing features — which encrypt data while it’s being processed — to protect user data in third-party data centers. Nvidia markets this tech as a way to keep data encrypted during computation on supported hardware. Features and scope - The redesign is being billed as Siri’s largest overhaul since 2011. Apple initially previewed new capabilities in 2024, including stronger personal context awareness, the ability to understand what’s on a user’s iPhone or iPad screen, and possibly a dedicated app positioned to compete with ChatGPT and other assistants. Reports also say Siri may integrate into the iPhone’s Dynamic Island with a new gesture-driven interface. Industry signals - The report echoes an earlier 2025 rumor that Apple ordered about 250 Nvidia NVL72 servers at roughly $4 million each. Nvidia’s Blackwell B200 architecture, introduced in 2024, is aimed at large language model training and inference — precisely the sorts of workloads Apple appears to need. What this means for crypto and decentralization - For crypto and web3 watchers, Apple’s turn to Nvidia and Google highlights the centralization of cutting-edge AI infrastructure: the fastest inference chains currently live on specialized, vendor-controlled hardware in hyperscale clouds. Apple’s use of confidential computing could ease privacy worries, but it also underscores opportunities for decentralized compute and privacy-preserving cloud alternatives to capture demand from big AI clients seeking trust-minimized processing. Status: unconfirmed - Apple has not publicly confirmed the reported arrangement. The WWDC keynote on June 8 should clarify how far Apple has come in rebuilding Siri after nearly two years of delays.